JP2006065584A - Abnormality detection apparatus - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an abnormality detection apparatus capable of accurately detecting the existence of abnormality in a caution area. <P>SOLUTION: The abnormality detection apparatus 1 is provided with a distance image sensor part 2 for photographing a target space Y and generating a distance image in which a distance value is set as a pixel value and a signal processing part 10 as main constitution. The signal processing part 10 is constituted of an image input means 11 for inputting the image data of the distance image from the distance image sensor part 2, a caution area setting means 12 for setting an area including at least a caution target X as a caution area in the distance image, an abnormality judging means 13 for judging the occurrence of abnormality in the caution target X when a change is generated in the image of the caution area in the distance image inputted to the image input means 11 and then outputting an audible signal, and an alarm means 14 for generating an alarm when the audible signal is inputted from the abnormality judging means 13. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to an abnormality detection device that detects the absence of an abnormality in a warning area.

  2. Description of the Related Art Conventionally, there has been provided a detection device that detects movement of a vehicle within a monitoring area by capturing a grayscale image of the monitoring area and performing pattern matching with a template image (see, for example, Patent Document 1).

An anomaly detection device that detects an anomaly in the area to be alerted using such pattern matching technology has also been proposed in the past. For example, the inside of a building to be alerted is imaged with a TV camera, and the shading imaged with the TV camera is captured. An image was image processed to detect abnormalities inside the building.
JP-A-11-284997

  In the above-described abnormality detection device, by detecting a change between a grayscale image of the space to be alerted (inside the building) captured by the TV camera and an image of the alert area (template image) captured at normal time, Although the presence / absence of an abnormal state is determined, since the gray image in the space to be alerted changes depending on the illumination condition, there is a possibility that a change in the illumination condition is erroneously detected as the occurrence of the abnormal state.

  The present invention has been made in view of the above problems, and an object of the present invention is to provide an abnormality detection device that can accurately detect the presence or absence of an abnormality in a warning area.

  In order to achieve the above object, the invention according to claim 1, a distance image sensor unit that captures an image of a target space and generates a distance image having a distance value as a pixel value, and a distance image input from the distance image sensor unit. A warning area setting unit that sets an area including at least a warning object as a warning area, and an abnormality determination unit that outputs a notification signal if there is a change in the image in the warning area in the distance image input from the distance image sensor unit And an alarm output unit that outputs an alarm when a notification signal is input.

  According to a second aspect of the present invention, in the first aspect of the present invention, the warning area setting unit has means for detecting a plane that is a placement surface in the distance image, and an object placed on the detected plane is detected. It is determined that the object is a warning object, and an area corresponding to the warning object is set as a warning area.

  According to a third aspect of the present invention, in the first aspect of the present invention, the alert area setting unit has means for designating a pixel corresponding to the alert object in the distance image, and the pixel value is approximately the center of the designated pixel. A continuous area is set as a warning area.

  According to a fourth aspect of the present invention, in the invention according to any one of the first to third aspects, the abnormality determination unit does not immediately output a warning signal when there is a change in the image of the warning area, but when there is a change. The output of the warning signal is stopped if the image of the warning area returns to the image before the change until a predetermined time elapses from the first time, and the warning signal is output if the image does not return to the image before the change. .

  The invention according to claim 5 is the invention according to any one of claims 1 to 4, wherein the abnormality determination unit does not immediately output a warning signal when there is a change in the image of the warning area, and the shape of the warning object. If the ratio of the portion where the change is not more than a certain ratio, the output of the reporting signal is stopped, and if the ratio is less than the certain ratio, the reporting signal is output.

  According to a sixth aspect of the present invention, in any one of the first to fifth aspects, the abnormality determination unit includes a means for storing a distance image of a warning object imaged in a normal state as a template image. The position when the correlation value is equal to or higher than the predetermined reference value by performing pattern matching by shifting the position of the template image relative to the image of the warning area without immediately outputting the alarm signal when there is a change When the determination condition that the deviation is smaller than a predetermined threshold is satisfied, the output of the notification signal is stopped, and when the determination condition is not satisfied, the notification signal is output.

  According to a seventh aspect of the present invention, in any one of the first to fifth aspects, the abnormality determining unit stores a distance image of a warning object captured during normal operation as a template image, and an edge pixel from the template image. Means for extracting, and means for extracting one or a plurality of blocks including a certain number of edge pixels as feature blocks, and the warning area is not output immediately when there is a change in the image of the warning area. Block matching is performed by relatively shifting the position of the feature block with respect to the image of the image, and the positional deviation when the correlation value exceeds a predetermined reference value for each feature block is smaller than a predetermined threshold value. If the determination condition is satisfied, the output of the alarm signal is stopped, and if the above determination condition is not satisfied, the alarm signal is output.

  The invention of claim 8 is characterized in that, in the invention of any one of claims 1 to 7, the alert area setting unit sets a range of a certain distance from the alert object as the alert area.

  According to the first aspect of the present invention, if there is a change in the image in the alert area in the distance image, the abnormality determining unit determines that some abnormality has occurred in the alert object and outputs a notification signal. The alarm output unit outputs an alarm according to the report signal, and the distance image input from the distance image sensor unit does not change the pixel value due to illumination fluctuation like the grayscale image, so the change of the illumination condition is abnormal It is possible to prevent erroneous detection.

  According to the invention of claim 2, the warning area setting unit detects a plane that is a placement surface in the distance image, determines that the object placed on the plane is a warning object, and Since the area corresponding to the object is set as the warning area, the work for the user to set the warning area can be reduced.

  According to the invention of claim 3, when the pixel corresponding to the alert target is designated in the distance image, the alert area setting unit alerts the area where the pixel values are substantially continuous around the designated pixel. It is assumed that the pixel corresponding to the alert object is set as an area, and the pixel values of the pixels corresponding to the alert object are assumed to be substantially continuous between adjacent pixels. The entire object can be set as a warning area, and the work for the user to set the warning area can be reduced.

  By the way, when another object moves in front of the alert object, the alert object is hidden behind the other object, and a change occurs in the image of the alert area. The unit does not immediately output a warning signal when there is a change in the image of the alert area, and the image in the alert area can return to the image before the change from when the change occurred until a predetermined time elapses. If it does not return to the pre-change image, it will output a warning signal, so it is abnormal if the warning object is temporarily hidden by another object moving on the near side. It is possible to prevent erroneous detection of occurrence of a state.

  Further, when a part of the alert object is hidden by another object that appears on the near side of the alert object, the image of the alert object is partially changed. The abnormality determination unit does not immediately output a warning signal when there is a change in the image of the warning area, and if the ratio of the part where the shape of the warning object has not changed is greater than a certain percentage, Since the output is stopped and the alarm signal is output if the ratio is less than a certain ratio, the alert object is hidden behind another object if the shape of the alert object does not change more than a certain ratio. Therefore, it is possible to prevent erroneous detection as an abnormal state.

  Furthermore, when the position of the warning object is slightly shifted, a change occurs in the image of the warning area. According to the invention of claim 6, the abnormality determination unit uses the distance image of the warning object captured during normal operation as a template. With a means for storing as an image, when there is a change in the image of the alert area, the warning signal is not output immediately, and the pattern image is compared with the position of the template image relative to the image of the alert area, If the determination condition that the positional deviation when the correlation value is equal to or greater than the predetermined reference value is smaller than the predetermined threshold is satisfied, the output of the alarm signal is stopped. If the determination condition is not satisfied, the alarm is output. Since the signal is output, the false alarm can be prevented when the position of the alert object is slightly shifted and the shape thereof is not changed.

  Furthermore, when the entire orientation of the alert object is slightly shifted, or when the posture of the person is changed when the alert object is a human, the edge pixel such as the corner of the alert object or the human head or waist is changed. According to the invention of claim 7, the abnormality determination unit includes means for storing a distance image of a warning object captured during normal operation as a template image, and a template. A means for extracting edge pixels from an image and a means for extracting one or more blocks including at least a certain number of edge pixels as feature blocks, and immediately outputting a warning signal when there is a change in the image of the alert area Without matching, block matching is performed by relatively shifting the position of the feature block with respect to the image of the warning area, and the positional deviation when the correlation value is greater than or equal to a predetermined reference value for each feature block If the condition that the threshold value is smaller than the predetermined threshold is satisfied, the output of the alarm signal is stopped, and if the determination condition is not satisfied, the alarm signal is output, so the position of the feature block is slightly shifted. If the shape does not change, false alarms can be prevented.

  According to the invention of claim 8, the warning area setting unit sets a range of a certain distance from the warning object as the warning area, and the abnormality determination unit outputs a notification signal if there is a change in the image of the warning area. Since it is outputting, there exists an effect that it can detect and alert | report that the other object entered into the inside of the warning area set to the range of the fixed distance from the warning target object.

  Embodiments of the present invention will be described below with reference to the drawings.

  FIG. 1 is a block diagram of an embodiment of an abnormality detection apparatus. The abnormality detection apparatus 1 includes a distance image sensor unit 2 and a signal processing unit 10 as main components. The abnormality detection device 1 according to the present embodiment is used for monitoring the safety of a warning object (for example, a safe) X installed on the ceiling 101 of the room 100 and placed in the room 100, for example.

  First, the configuration of the distance image sensor unit 2 used in the present embodiment will be described. FIG. 2 shows a block diagram of the distance image sensor unit 2. The distance image sensor unit 2 includes a light source 5 that irradiates light to a target space Y in which a warning object X exists, and light from the target space Y. And a light detection element 3 that can obtain an electrical output of an output value reflecting the amount of received light. The distance from the light source 5 to the target space Y is applied to the target object in the target space Y. And the time from when the reflected light from the object enters the light detection element 3 (referred to as “flight time”). However, since the flight time is very short, the intensity modulated light that is modulated so that the intensity of the light irradiating the target space Y periodically changes at a constant period is used, and the phase when the intensity modulated light is received is used. Seeking flight time.

  That is, as shown in FIG. 10A, it is assumed that the intensity of light radiated from the light emitting source 5 into the space changes as shown by curve A, and the amount of received light received by the light detecting element 3 changes as shown by curve B. For example, since the phase difference ψ corresponds to the flight time, the distance to the object can be obtained by obtaining the phase difference ψ. Further, the phase difference ψ can be calculated using the received light quantity of the curve B obtained at a plurality of timings of the curve A. For example, it is assumed that the received light amounts of curve B obtained with the phases of curve A at 0 degrees, 90 degrees, 180 degrees, and 270 degrees are A0, A1, A2, and A3, respectively. However, the received light amount A0, A1, A2, A3 in each phase uses the received light amount integrated at a predetermined time Tw instead of the instantaneous value. Now, while obtaining the received light amounts A0, A1, A2, and A3, the phase difference ψ does not change (that is, the distance to the object does not change), and the reflectance of the object does not change. To do. Further, it is assumed that the intensity of light emitted from the light emitting source 5 is modulated with a sine wave, and the intensity of light received by the light detection element 3 at time t is represented by A · sin (ωt + δ) + B. Here, A is the amplitude, B is the external light component, ω is the angular frequency, and δ is the phase. If the received light amounts A0, A1, A2, A3 received by the light detecting element 3 are instantaneous values instead of the integrated values of the time Tw, the received light amounts A0, A1, A2, A3 can be expressed as follows.

A0 = A · sin (δ) + B
A1 = A · sin (π / 2 + δ) + B
A2 = A · sin (π + δ) + B
A3 = A · sin (3π / 2 + δ) + B
Since δ = −ψ, A0 = −A · sin (ψ) + B, A1 = A · cos (ψ) + B, A2 = A · sin (ψ) + B, A3 = −A · cos (ψ ) + B, and as a result, the relationship between the received light amounts A0, A1, A2, A3 and the phase difference ψ is expressed as follows.

ψ = tan ⁻¹ {(A2−A0) / (A1−A3)} (1)
In the above equation (1), the instantaneous values of the received light amounts A0, A1, A2, A3 are used, but even if the integrated values at the time Tw are used as the received light amounts A0, A1, A2, A3, the equation (1) ) To obtain the phase difference ψ.

  By the way, in order to modulate the intensity of the light irradiating the target space Y, as the light emitting source 5, for example, a structure in which a large number of light emitting diodes are arranged on one plane or a combination of a semiconductor laser and a diverging lens is used. Use. The light emission source 5 is driven by a modulation signal having a predetermined modulation frequency output from the control circuit unit 7, and the intensity of the light emitted from the light emission source 5 is modulated by the modulation signal. The control circuit unit 7 modulates the intensity of light emitted from the light source 5 with, for example, a 20 MHz sine wave. The intensity of the light emitted from the light source 5 may be modulated by a triangular wave, a sawtooth wave, or the like in addition to the modulation by a sine wave. In short, any configuration is possible as long as the intensity is modulated at a constant period. May be adopted.

  The light detecting element 3 includes a plurality of photosensitive portions 3a regularly arranged. A light receiving optical system 4 is disposed in the light incident path to the photosensitive portion 3a. The photosensitive portion 3a is a portion where light from the target space Y enters through the light receiving optical system 4 in the light detecting element 3, and the photosensitive portion 3a generates an amount of electric charge corresponding to the amount of received light. Further, the photosensitive portions 3a are arranged on lattice points of a planar lattice, and are arranged in a matrix in which, for example, a plurality are arranged at equal intervals in the vertical direction (that is, the longitudinal direction) and the horizontal direction (that is, the lateral direction). Is done.

  The light receiving optical system 4 associates the line-of-sight direction when viewing the target space Y from the light detection element 3 with each photosensitive portion 3a. That is, the range in which light is incident on each photosensitive portion 3a through the light receiving optical system 4 can be regarded as a conical field of view having a small apex angle set for each photosensitive portion 3a with the center of the light receiving optical system 4 as a vertex. it can. Therefore, if the reflected light emitted from the light source 5 and reflected by the object existing in the target space Y enters the photosensitive portion 3a, the optical axis of the light receiving optical system 4 depends on the position of the photosensitive portion 3a that has received the reflected light. It is possible to know the direction in which the object exists with reference to.

  Since the light receiving optical system 4 is generally arranged so that the optical axis is orthogonal to the plane on which the photosensitive portion 3a is arranged, the center of the light receiving optical system 4 is set as the origin, and the vertical and horizontal directions of the plane on which the photosensitive portion 3a is arranged If an orthogonal coordinate system is set in which the optical axis of the light receiving optical system 4 is in the direction of three axes, the angle (so-called azimuth and elevation) when the position of the target existing in the target space Y is expressed in spherical coordinates. It corresponds to each photosensitive portion 3a. The light receiving optical system 4 can also be arranged so that the optical axis intersects at an angle other than 90 degrees with respect to the plane on which the photosensitive portions 3a are arranged.

  In the present embodiment, as described above, in order to obtain the distance to the object, the received light amounts A0, A1, and A2 are synchronized at four timings synchronized with the intensity change of the light emitted from the light source 5 to the target space Y. , A3. Therefore, it is necessary to control the timing to obtain the desired received light amount A0, A1, A2, A3. Further, since the amount of charge generated in the photosensitive portion 3a is small in one cycle of the intensity change of light irradiated from the light source 5 to the target space Y, it is desirable to accumulate the charges over a plurality of cycles. Therefore, as shown in FIG. 2, by providing a plurality of charge accumulating portions 3c for accumulating the charges generated in each photosensitive portion 3a, and changing the area of a region for generating charges that can be used in each photosensitive portion 3a. A plurality of sensitivity control units 3b for adjusting the sensitivity of each photosensitive unit 3a are provided.

  In each sensitivity control unit 3b, the sensitivity of the photosensitive unit 3a corresponding to the sensitivity control unit 3b is increased at any one of the four points described above, and in the photosensitive unit 3a with increased sensitivity, the received light amount A0, Since charges corresponding to A1, A2, and A3 are mainly generated, charges corresponding to the received light amounts A0, A1, A2, and A3 can be accumulated in the charge accumulating portion 3c corresponding to the photosensitive portion 3a.

  By the way, the sensitivity control unit 3b changes the amount of charge generated in each period by changing the area (substantial light receiving area) of the region that generates charge that can be used in the photosensitive unit 3a. The charges accumulated in 3c are not necessarily mixed with the charges generated in the period in which the received light amounts A0, A1, A2, A3 are obtained, but also the charges generated in other periods. Now, in the sensitivity control unit 3b, the sensitivity in the period for generating the charges corresponding to the received light amounts A0, A1, A2, A3 is α, the sensitivity in the other periods is β, and the photosensitive unit 3a is a charge proportional to the received light amount. Is generated. Under this condition, the charge accumulating unit 3c that accumulates charges corresponding to the received light amount A0 is proportional to αA0 + β (A1 + A2 + A3) + βAx (Ax is the received light amount other than the period during which the received light amounts A0, A1, A2, and A3 are obtained). Charges proportional to αA2 + β (A0 + A1 + A3) + βAx are accumulated in the charge accumulating unit 3c that accumulates charges and accumulates charges corresponding to the received light quantity A2. As described above, when obtaining the phase difference ψ, (A2−A0) is obtained, and A2−A0 = (α−β) (A2−A0), and similarly, A1−A3 = (α−). β) Since (A1-A3), (A2-A0) / (A1-A3) theoretically has the same value regardless of the presence or absence of charge mixing. The phase difference ψ has the same value.

  The photodetecting element 3 including the photosensitive unit 3a, the sensitivity control unit 3b, and the charge accumulating unit 3c is configured as one semiconductor device, and the photodetecting element 3 receives charges accumulated in the charge accumulating unit 3c outside the semiconductor device. A charge extraction portion 3d is provided for extraction. The charge extraction unit 3d has the same configuration as the vertical transfer unit and horizontal transfer unit in the CCD image sensor.

  As described above, since each photosensitive unit 3a generates an amount of electric charge corresponding to the amount of received light, each of the received light amounts A0, A1, A2, A3 reflects the brightness of the object. That is, the added value or average value of the received light amounts A0, A1, A2, A3 corresponds to the density value in the grayscale image. In other words, in addition to obtaining the distances from the received light amounts A0, A1, A2, A3 to the object at each photosensitive portion 3a, it is also possible to obtain the density value of the object. In addition, since the distance and density value of the object are obtained using the photosensitive portion 3a at the same position, information on both the density value and the distance can be obtained for the same position.

  The charges extracted from the charge extraction unit 3d in this way are converted into digital quantities by the A / D conversion unit 9, and then given as image signals to the image generation unit 8 made of DSP. The distance to the object in Y is calculated from the received light amounts A0, A1, A2, and A3 using the above formula. That is, the image generation unit 8 calculates the distance to the object in each direction corresponding to each photosensitive unit 3a, and calculates the three-dimensional information of the target space Y. By using this three-dimensional information, it is possible to generate a distance image in which the pixel values of the pixels matching in each direction of the target space Y are distance values. Further, the image generation unit 8 generates a grayscale image of the target space Y based on the density value obtained by each photosensitive unit 3a, and this grayscale image uses the average value of the received light amounts A0, A1, A2, A3 as the grayscale. If it is used for the value, the influence of light from the light source 5 can be removed. That is, the image generation unit 8 generates a grayscale image and a distance image. In this embodiment, the image generation unit 8 and the control circuit unit 7 are configured by a one-chip ASIC 6.

  Then, the distance image and grayscale image of the target space Y obtained by the image generation unit 8 of the distance image sensor unit 2 are input to the signal processing unit 10, and the presence or absence of abnormality in the target space Y is detected by the signal processing unit 10. Processing is performed.

  The signal processing unit 10 is composed of, for example, a microcomputer, and includes an image input means 11, a warning area setting means 12, an abnormality determination means 13, and an alarm means 14. The warning area setting means 12 and the abnormality determination means 13 are computing functions of the microcomputer. It is realized by.

  The image input means 11 has an image memory (not shown), and temporarily stores the distance image and grayscale image data input from the distance image sensor unit 2 in the image memory and also stores the image data stored in the image memory. Is used to display a distance image and a grayscale image of the target space Y on a display means (not shown) such as a TV monitor.

  The warning area setting means 12 includes an input device such as a mouse or a pen input device, and the security officer operates the input device within the distance image displayed on the display means to specify the warning area. Is set (for example, a collection of pixels corresponding to the alert object X). Here, when a warning area is set using the warning area setting means 12, it is included in the surface (for example, floor surface) on which the warning object X is placed in the distance image of the target space Y as shown in FIG. Three reference points P1, P2, and P3 are input to the security officer, and the warning area setting means 12 applies the three reference points P1, P2, and P3 to the plane, and the distance to the point on the plane and the pixel A pixel having a small value difference is determined to be a pixel corresponding to a point included in the plane, and a pixel having a large pixel value difference is determined to be a warning target object X placed on the plane. The pixel corresponding to the object X is set as a warning area, and the security officer can set the pixel corresponding to the warning object X as the warning area only by designating three points on the plane as the placement surface. Therefore, the setting work of the alert area can be reduced.

  Note that instead of having the security officer input the reference points P1, P2, and P3 on the plane serving as the placement surface, the warning area setting unit 12 is an area of pixels in which distance values are substantially continuous in the distance image of the target space Y. An area of a certain area or more is determined as a plane (floor surface) serving as a placement surface, and a pixel having a large difference in distance value from this floor surface is determined as a warning object X placed on the floor surface. The pixels corresponding to the alert object X may be set in the alert area, and the alert area setting operation can be eliminated.

  Also, as shown in FIG. 4, the security officer is caused to input a point P4 corresponding to the warning object X in the distance image of the target space Y, and the warning area setting means 12 is in the vicinity of 8 pixels of the pixel corresponding to the point P4. The pixels having a small difference in pixel value (distance value) are grouped as the same group, and the difference between the pixel values of each of the eight neighboring pixels of the point P4 is compared with the neighboring eight neighboring pixels. By grouping pixels into the same group and continuing this process, an area where pixel values are substantially continuous with the point P4 as the center is extracted as a warning object X, and the extracted warning object X is set as a warning area. In other words, it is possible to extract a pixel corresponding to the alert object X and set it in the alert area by only specifying one point of the alert object X.

  The abnormality determination unit 13 performs a process for determining whether or not there is an abnormality in the warning area set by the warning area setting unit 12 based on the distance image input to the image input unit 11, and when an abnormality is detected, the alarm signal Is output.

  The alarm means 14 is composed of, for example, a buzzer or a lamp. When an alarm signal is input from the abnormality determination means 13, the alarm person is informed of the occurrence of an abnormality by sounding the buzzer or blinking the lamp. To do. Note that the alarm means 14 may be provided with a communication means for outputting a notification signal to an external device, and the occurrence of an abnormality can be notified to a device at another location.

  Next, the operation of the signal processing unit 10 having the above configuration will be described below.

  The distance image sensor unit 2 irradiates the target space Y with intensity-modulated light from the light source 5, and receives the reflected light reflected by the target object in the target space Y by the light detection element 3. A distance image and a gray image are generated.

  In the signal processing unit 10, the image input unit 11 captures the distance image and the grayscale image data from the distance image sensor unit 2 at predetermined time intervals, stores the captured image data in the image memory, and displays the image data on the screen of the display unit. To display.

  Here, when the warning area is not set such as when the power is turned on or when the warning is started, the security officer performs the warning area setting operation using the input means. For example, when the security officer designates a pixel corresponding to the alert object X using the input device provided in the alert area setting unit 12 while viewing the distance image of the target space Y displayed on the display unit, the designated pixel An area in which pixel values are substantially continuous with the center of is set as a warning area. When the warning area is set by the warning area setting means 12, the abnormality determination means 13 generates the warning object X from the distance image generated by the distance image sensor unit 2 in a normal state and held in the image input means 11. The image of the alert area is cut out, and the cut out image is registered in a storage unit (not shown) as a template image.

  Then, when the image input unit 11 captures the distance image from the distance image sensor unit 2 at a predetermined timing during the warning operation, the abnormality determination unit 13 cuts out the image of the warning area from the distance image held in the image input unit 11. Then, template matching is performed between the cut-out image and the template image described above to obtain a correlation value (degree of coincidence) between the two images, and when the correlation value becomes smaller than a predetermined threshold (that is, the alert object X When the distance image greatly changes from the template image), it is determined that an abnormality has occurred in the alert object X, and an alarm signal is output. At this time, an alarm signal is input from the abnormality determination unit 13 to the alarm unit 14, and the alarm unit 14 outputs an alarm, so that the safety staff can be notified of the occurrence of the abnormality. The abnormality determination unit 13 performs matching processing with the template image for the image of the warning area cut out from the distance image, but performs matching processing for the image of the warning area cut out from the contour image extracted from the distance image. May be.

  Here, as shown in FIG. 5, when another object (for example, a person M) appears on the front side of the alert object X, a part of the alert object X is hidden behind the person M, and the alert object X However, if the alert object X is simply concealed by another object on the near side and the shape of the alert object X is substantially the same, the alarm signal is not output. Is preferred. Therefore, the abnormality determination unit 13 cuts out the image of the warning area from the distance image held in the image input unit 11, performs template matching between the cut out image and the template image, and correlates the two images. As a result, when the correlation value becomes lower than a predetermined threshold value, the abnormality determination means 13 does not immediately output a warning signal, but a portion where the shape of the alert object has not changed (that is, the pixel If the ratio of pixels whose values have not changed is equal to or greater than a certain ratio, the output of the alarm signal may be stopped, and if the ratio is less than a certain ratio, the alarm signal may be output. If only a part of the object X is concealed by another object and the position or shape of the alert object X has not changed, the output of the alarm signal is stopped, thereby preventing false alarms. .

  Further, as shown in FIGS. 6A to 6C, when the person M passes the front side of the alert object X, the alert object X is temporarily hidden behind the person M as shown in FIG. Since the image of the warning object X changes because it is hidden, even if there is no change in the position or shape of the warning object X before and after the person M passes as shown in FIGS. An alarm will be output. Therefore, when the correlation value obtained by pattern matching performed by the abnormality determination unit 13 becomes lower than a predetermined threshold value, the abnormality determination unit 13 does not immediately output an alarm signal, and the correlation value exceeds the predetermined threshold value. If the correlation value becomes greater than or equal to the threshold value (after returning to the image before the change) from when the value becomes low until the predetermined time elapses, the alarm signal is not output and the correlation value does not exceed the threshold value. (If the image does not return to the image before the change), a notification signal may be output when a predetermined time has elapsed, and it is possible to prevent erroneous detection of temporary concealment as the occurrence of an abnormality.

  In addition, as shown in FIG. 7, when the position of the alert object X is slightly moved from the position indicated by the dotted line to the position indicated by the solid line in the figure, the pixel value of the pixel corresponding to the alert area changes. There is a possibility that the correlation value between the image of the region and the template image is lowered and an alarm is output from the alarm means 14, but if there is no change in the shape of the alarm object X, such a minute position It is preferable not to output an alarm at the deviation. Therefore, the abnormality determination unit 13 cuts out the image of the warning area from the distance image held in the image input unit 11, performs template matching between the cut out image and the template image, and correlates the two images. As a result, when the correlation value is lower than a predetermined threshold value, the abnormality determination unit 13 does not immediately output the alarm signal, but the distance image near the warning area input from the image input unit 11 On the other hand, pattern matching is performed by relatively shifting the position of the template image in an arbitrary direction, and the determination that the positional deviation when the correlation value between the two is equal to or greater than a predetermined reference value is smaller than a predetermined threshold value. If the condition (1) is satisfied, it is determined that the position of the warning object X does not change and the position is slightly shifted, the output of the alarm signal is stopped, and the above determination is made. If matter (1) is satisfied, the abnormality is determined to have occurred, may be configured to output the alarm signal can be prevented from being erroneously detected small misalignment and abnormal. As a case where the above-described determination condition (1) is not satisfied, the positional deviation of the alert object X is large, and the positional deviation when the correlation value is equal to or greater than the reference value is larger than a predetermined threshold value. Alternatively, there may be a case where the alert object X moves further and there is no partial image whose correlation value is equal to or greater than the reference value.

  In addition, as shown in FIG. 8, when the direction of the warning object X is slightly shifted from the direction indicated by the dotted line in the drawing to the direction indicated by the solid line, the pixel value of the pixel corresponding to the warning area changes. There is a possibility that the correlation value between the image and the template image decreases and an alarm is output from the alarm means 14, but if there is no change in the shape or position of the alarm object X, such a minute It is preferable not to output an alarm at an angle shift. Here, when the whole direction of the alert object X is slightly shifted or the posture of the person who is the alert object is changed, the corner of the alert object X or the human being when the alert object is a human being It is considered that the image of the feature block including many edge pixels such as the head and waist is slightly shifted. Therefore, the abnormality determination unit 13 cuts out the image of the warning area from the distance image held in the image input unit 11, performs template matching between the cut-out image and the template image, and correlates the two images. As a result, when the correlation value becomes lower than a predetermined threshold value, the abnormality determination means 13 does not immediately output a notification signal, but performs edge extraction processing on the template image in advance to make the edge pixels constant. Blocks including the above (for example, 3 × 3 pixels, 5 × 5 pixels) are extracted as feature blocks B1 to B4, and block matching is performed by relatively shifting the positions of the feature blocks B1 to B4 with respect to the image of the alert area. The determination condition (2) that the positional deviation when the correlation value exceeds a predetermined reference value for each feature block B1 to B4 is smaller than a predetermined threshold value. ) Is satisfied, the shape and position of the warning object X do not change, and it is determined that a slight angle deviation has occurred, and the output of the alarm signal is stopped, and the above-described determination condition (2) is If it does not hold, it may be determined that the direction of the alert object X has changed greatly or the shape or position has changed, and a warning signal may be output. Can be prevented from being erroneously detected. In addition, when said determination condition (2) is not satisfied, the direction of the warning object X changes greatly, and the position shift when a correlation value becomes more than a reference value becomes larger than a predetermined threshold value. The case where the direction of the warning target object X changes greatly further and the block whose correlation value becomes more than a reference value does not exist is considered.

  As described above, in the present embodiment, whether the distance image of the alert area changes by performing template matching or inter-frame difference on the distance image of the alert area (alert object X) captured by the distance image sensor unit 2. If the distance image of the alert area changes, it can be determined that some change has occurred in the alert object X. Further, the abnormality determination means 13 determines the presence or absence of abnormality based on the distance image of the warning area, and the distance image does not change due to the illumination fluctuation of the warning area unlike the grayscale image. It is possible to prevent erroneous detection of occurrence.

  The warning area setting means 12 extracts pixels corresponding to the warning object X from the distance image of the target space Y, and sets the extracted pixel aggregate as a warning area. A distance range may be set as a warning area, and by detecting a change in the distance image of the warning area, it may be detected that another object has entered a range of a certain distance from the warning object X. Here, as a method of setting a range of a certain distance from the alert object X as the alert area, for example, when the alert area setting unit 12 extracts a pixel corresponding to the alert object X, 3 × 3 from the extracted pixel is used. Planes M1, M2, and M3 are extracted by extracting a plane of a pixel or a small area of about 5 × 5 pixels and translating the plane to a position away from the alert object X in the normal direction by a certain distance. The space surrounded by these planes M1, M2, and M3 may be set as the alert area X0, and the alert area X0 can be automatically set using the detection result of the alert object X.

  Next, a specific structural example of the light detection element 3 used in the present embodiment will be described with reference to FIGS. The photodetecting element 3 shown in FIG. 11 has a plurality of (for example, 100 × 100) photosensitive portions 3a arranged in a matrix, and is formed on, for example, a single semiconductor substrate. In each column in the vertical direction of the photosensitive portion 3a, the semiconductor layer 21 that is integrally continuous is shared, and the semiconductor layer 21 is used as a transfer path for charges in the vertical direction (using electrons in this embodiment). It is possible to employ a configuration in which a semiconductor substrate is provided with a horizontal transfer portion Th that is a CCD that receives charges from one end of the semiconductor layer 21 and transfers the charges in the horizontal direction.

  That is, as shown in FIG. 12, the semiconductor layer 21 is used as both the photosensitive portion 3a and the charge transfer path, and is similar to a frame transfer (FT) type CCD image sensor. Similarly to the FT-type CCD image sensor, a light-shielded storage region Db is provided adjacent to the imaging region Da in which the photosensitive portions 3a are arranged, and charges accumulated in the storage region Db are transferred to the horizontal transfer unit Th. To do. The transfer of charges from the imaging area Da to the storage area Db is performed at once in the vertical blanking period, and the horizontal transfer unit Th transfers charges for one horizontal line in one horizontal period. The charge extraction unit 3d illustrated in FIG. 2 represents a function including a horizontal transfer unit Th as well as a function as a charge transfer path in the vertical direction in the semiconductor layer 21. However, the charge accumulation unit 3c does not mean the accumulation region Db, but represents a function of accumulating charges in the imaging region Da. In other words, the accumulation region Db is included in the charge extraction unit 3d.

  The semiconductor layer 21 is doped with impurities, the main surface of the semiconductor layer 21 is covered with an insulating film 22 made of an oxide film, and a plurality of control electrodes 23 are arranged on the semiconductor layer 21 via the insulating film 22. . This light detection element 3 has a structure known as a MIS element, but a normal MIS element is that a plurality of (five in the illustrated example) control electrodes 23 are provided in a region functioning as one light detection element 3. Is different. A material is selected for the insulating film 22 and the control electrode 23 so that light having the same wavelength as the light irradiated from the light source 5 to the target space can be transmitted. When light enters the semiconductor layer 21 through the insulating film 22, the semiconductor layer 21. A charge is generated inside the. The conductivity type of the semiconductor layer 21 in the illustrated example is n-type, and electrons e are used as charges generated by light irradiation. FIG. 11 shows only the region corresponding to one photosensitive portion 3a. As described above, a plurality of regions having the structure of FIG. 11 are arranged on the semiconductor substrate (not shown) and the charge extraction is performed. A structure to be the part 3d is provided. The vertical transfer unit provided as the charge extraction unit 3d is assumed to transfer charges in the left-right direction in FIG. 11, but it is also possible to adopt a configuration in which charges are transferred in a direction orthogonal to the plane in FIG. is there. In addition, when transferring charges in the horizontal direction in the figure, it is desirable to set the width dimension of the control electrode 23 in the horizontal direction to about 1 μm.

  In the photodetecting element 3 having this structure, when a positive control voltage + V is applied to the control electrode 23, a potential well (depletion layer) 24 that accumulates electrons e in a portion corresponding to the control electrode 23 is formed in the semiconductor layer 21. The That is, when light is applied to the semiconductor layer 21 with a control voltage applied to the control electrode 23 so as to form the potential well 24 in the semiconductor layer 21, a part of the electrons e generated in the vicinity of the potential well 24. Are captured in the potential well 24 and accumulated in the potential well 24, and the remaining electrons e disappear due to recombination in the deep part of the semiconductor layer 21. Further, the electrons e generated at a location away from the potential well 24 are also extinguished by recombination in the deep part of the semiconductor layer 21.

  Since the potential well 24 is formed at a portion corresponding to the control electrode 23 to which the control voltage is applied, the potential well 24 along the main surface of the semiconductor layer 21 is changed by changing the number of the control electrodes 23 to which the control voltage is applied. (In other words, the area of a region that generates a charge that can be used on the light receiving surface) can be changed. That is, changing the number of control electrodes 23 to which the control voltage is applied means adjusting sensitivity in the sensitivity control unit 3b. For example, when the control voltage + V is applied to three control electrodes 23 as shown in FIG. 11A, and when the control voltage + V is applied to one control electrode 23 as shown in FIG. The area occupied by the potential well 24 on the light receiving surface changes, and the area of the potential well 24 is larger in the state shown in FIG. As a result, the ratio of the charge that can be used increases and the sensitivity of the photosensitive portion 3a is substantially increased. Thus, it can be said that the photosensitive portion 3 a and the sensitivity control portion 3 b are constituted by the semiconductor layer 21, the insulating film 22, and the control electrode 23. Since the potential well 24 holds charges generated by light irradiation, it functions as the charge accumulating portion 3c.

  In order to extract charges from the potential well 24, a technique similar to that of the FT type CCD may be employed. After the electrons e are accumulated in the potential well 24, a control voltage of an applied pattern different from that during charge accumulation is controlled. By applying the voltage to the electrode 23, the electrons e accumulated in the potential well 24 can be transferred in one direction (for example, the right direction in the figure). That is, the semiconductor layer 21 can be used as a charge transfer path in the same manner as the vertical transfer portion of the CCD. Further, the electric charge is transferred through the horizontal transfer unit Th shown in FIG. 12 and is taken out of the photodetecting element 3 from an electrode (not shown) provided on the semiconductor substrate. In short, by controlling the application pattern of the control voltage to the control electrode 23, the sensitivity of each photosensitive portion 3a is controlled, charges generated by light irradiation are integrated, and the integrated charges are transferred. Can do.

  The sensitivity control unit 3b according to the present embodiment switches the sensitivity of the photosensitive unit 3a between two levels, high and low, by switching the area for generating available charges into two levels, that is, the received light amounts A0, A1, A2, and A3. High sensitivity is set only during a period in which the charge corresponding to any of the photosensitive portions 3a is to be generated (the area for generating charges is increased), and low sensitivity is set in other periods. The period of high sensitivity and the period of low sensitivity are set in synchronization with the modulation signal for driving the light source 5. Further, after the charges are accumulated in the potential well 24 over a plurality of periods of the modulation signal, the charges are taken out of the light detection element 3 through the charge extraction portion 3d. The charge is accumulated over a plurality of periods of the modulation signal because the period for generating the usable charge in the photosensitive portion 3a is short within one period of the modulation signal (for example, if the frequency of the modulation signal is 20 MHz). This is because less than a quarter of 50 ns is generated. That is, by integrating charges for a plurality of periods of the modulation signal, signal charges (charges corresponding to light emitted from the light emission source 5) and noise charges (external light components and shots generated inside the light detection element 3). A large signal-to-noise ratio can be obtained, and a large SN ratio can be obtained.

  By the way, if the charges corresponding to the four kinds of received light amounts A0, A1, A2, and A3 necessary for obtaining the phase difference ψ are generated by one photosensitive portion 3a, the resolution with respect to the line-of-sight direction is increased. There arises a problem that the time difference for obtaining the charges corresponding to the received light amounts A0, A1, A2, A3 becomes large. On the other hand, if the charges corresponding to the received light amounts A0, A1, A2, A3 are respectively generated by the four photosensitive portions 3a, the time difference for obtaining the charges corresponding to the received light amounts A0, A1, A2, A3 becomes small. However, a shift occurs in the line-of-sight direction for obtaining the four types of charges, and the resolution in the line-of-sight direction decreases. Therefore, in the present embodiment, a configuration is adopted in which two types of charges corresponding to the received light amounts A0, A1, A2, and A3 are generated within one cycle of the modulation signal by using the two photosensitive portions 3a. . That is, two photosensitive portions 3a are used as a set, and light from the same line-of-sight direction is incident on the two photosensitive portions 3a.

  By adopting the above-described configuration, the resolution in the line-of-sight direction can be made relatively high, and the time difference for generating charges corresponding to the received light amounts A0, A1, A2, and A3 can be reduced. In other words, by reducing the time difference for generating charges corresponding to the received light amounts A0, A1, A2, and A3, the distance detection accuracy can be kept relatively high even for an object moving in the object space. Can do. In the configuration of this embodiment, the resolution in the line-of-sight direction is lower than that in the case where the charge corresponding to the four types of received light amounts A0, A1, A2, and A3 is generated by one photosensitive unit 3a. The resolution can be improved by downsizing the photosensitive part 3a or designing the light receiving optical system 4.

  The operation of the light detection element 3 will be specifically described below. The example shown in FIG. 11 shows an example in which five control electrodes 23 are provided for one photosensitive portion 3a. However, the two control electrodes 23 on both sides are charged (electrons e) by the photosensitive portion 3a. Is used to form a barrier for preventing the electric charge from flowing out to the adjacent photosensitive portion 3a, and when the two photosensitive portions 3a are used as a set, the adjacent photosensitive portion Since a barrier is formed by any one of the photosensitive portions 3a between the potential wells 24 of 3a, it is sufficient to provide three control electrodes 23 for each photosensitive portion 3a. With this configuration, the occupation area per one photosensitive portion 3a is reduced, and it is possible to suppress a decrease in resolution in the line-of-sight direction while using two photosensitive portions 3a as a set.

  Here, as shown in FIG. 13, the numbers (1) to (6) are assigned to the control electrodes 23 in order to distinguish the three control electrodes 23 respectively provided in the two photosensitive portions 3a. Attached. The two photosensitive portions 3a having the control electrodes 23 to which the numbers (1) to (6) are assigned correspond to one line-of-sight direction and constitute pixels in the image sensor. It is desirable to provide an overflow drain in association with the photosensitive portion 3a for each pixel.

  13A and 13B show a state in which a control voltage + V is applied to the control electrode 23 in a different application pattern (a state in which the control voltage + V is applied between a substrate electrode (not shown) provided on the semiconductor substrate and the control electrode 23). As can be seen from the shape of the potential well 24, a positive control voltage + V is applied to the control electrodes (1) to (3) of the two photosensitive portions 3a to be one pixel in FIG. In addition, a positive control voltage + V is applied to the central control electrode (5) among the remaining control electrodes (4) to (6). In (b), a positive control voltage + V is applied to the central control electrode (2) of the control electrodes (1) to (3), and positive to the remaining control electrodes (4) to (6). The control voltage + V is applied. That is, the application pattern of the control voltage + V applied to the two photosensitive portions 3a constituting one pixel is alternately replaced. The timing of switching the application pattern of the control voltage + V applied to the two photosensitive portions 3a is the timing of the opposite phase (the phase is 180 degrees different) in the modulation signal. In addition, except for the period in which the control voltage + V is simultaneously applied to the three control electrodes 23 provided in each photosensitive portion 3a, one central control electrode 23 (that is, the control electrode) provided in each photosensitive portion 3a. (2) The control voltage + V is applied only to (5)), and the other control electrodes 23 are kept at 0V.

  For example, when the charges corresponding to the received light amounts A0 and A2 are alternately generated in the two photosensitive portions 3a constituting one pixel, as shown in FIG. 13, one of the photosensitive portions 3a corresponds to the received light amount A0. While the control voltage + V is applied to the three control electrodes (1) to (3) in order to generate charges, the other photosensitive portion 3a has one to hold the charge corresponding to the received light quantity A2. A control voltage + V is applied only to the control electrode (5). Similarly, while the control voltage + V is being applied to the three control electrodes (4) to (6) in order to generate a charge corresponding to the received light amount A2 in one photosensitive portion 3a, the other photosensitive region is exposed. In the unit 3a, the control voltage + V is applied only to one control electrode (2) in order to hold the charge corresponding to the received light quantity A0. In addition, the control voltage + V is applied only to the control electrodes (2) and (5) except for the period in which charges corresponding to the received light amounts A0 and A2 are generated. FIGS. 10B and 10C show the application timing of the control voltage + V to the control electrodes (1) to (6) when accumulating charges corresponding to the received light amounts A0 and A2. In the figure, the hatched portion indicates a state where the control voltage + V is applied, and the blank portion indicates a state where no voltage is applied to the control electrodes (1) to (6).

  The same applies to the case where the charges corresponding to the received light amounts A1 and A3 are generated in the two photosensitive portions 3a constituting one pixel, and the case where the charges corresponding to the received light amounts A0 and A2 are generated is different from the control electrode 23. The only difference is that the timing at which the control voltage + V is applied differs by 90 degrees in the phase of the modulation signal. In addition, the charge is transferred from the imaging region to the accumulation region between a period for generating charges corresponding to the received light amounts A0 and A1 and a period for generating charges corresponding to the received light amounts A1 and A3. That is, charges corresponding to the received light amount A0 are accumulated in the potential well 24 corresponding to the control electrodes (1) to (3), and charges corresponding to the received light amount A2 correspond to the control electrodes (4) to (6). When accumulated in the potential well 24, the charges corresponding to the received light amounts A0 and A2 are taken out. Next, charges corresponding to the received light amount A1 are accumulated in the potential well 24 corresponding to the control electrodes (1) to (3), and charges corresponding to the received light amount A3 are applied to the control electrodes (4) to (6). When accumulated in the corresponding potential well 24, charges corresponding to these received light amounts A1 and A3 are taken out to the outside. By repeating such an operation, the charges corresponding to the received light amounts A0, A1, A2, and A3 in the four sections can be taken out of the light detecting element 3 by two reading operations, and the extracted charges are used. The phase difference ψ can be obtained. For example, in order to obtain an image of 30 frames per second, the period for generating charges corresponding to the received light amounts A0 and A1 and the period for generating charges corresponding to the received light amounts A1 and A3 are shorter than 1/60 second. A short period.

  In the above example, the control voltage applied simultaneously to the three control electrodes 23 ((1) to (3) or (4) to (6)) and one control electrode 23 ((2) or (5)) Since the control voltage applied only to is equal, the area of the potential well 24 changes, but the depth of the potential well 24 is equal. In this case, the charges generated at the control electrode 23 ((1) (3) or (4) (6)) to which no control voltage is applied flow into the potential well 24 with a similar probability. That is, by applying the control voltage + V to only one of the three control electrodes 23 constituting the photosensitive portion 3a, the region functioning as the charge accumulation portion 3c and all the three control electrodes 23 are applied. A similar amount of charge flows into both the region to which the control voltage + V is applied. That is, since the noise component flowing into the potential well 24 holding the charge is relatively large, the dynamic range is lowered.

  Therefore, as shown in FIG. 14, the control voltage applied simultaneously to each of the three control electrodes (1) to (3) or (4) to (6) provided in the two photosensitive portions 3a in the set is 1 It is set to be lower than the control voltage applied only to the individual control electrodes (2) or (5), and the depth of the small area potential well 24 is set to be smaller than the depth of the large area potential well 24. Is desirable. As described above, the potential well 24 that mainly generates charges (electrons e) is made deeper than the potential well 24 that mainly holds charges, so that the control electrode (1) to which no control voltage is applied is applied. Charges generated at the sites corresponding to (3) or (4) and (6) are likely to flow into the deeper potential well 24. That is, as compared with the case where a constant control voltage + V is applied to the control electrode 23, the noise component flowing into the potential well 24 holding the charge can be reduced.

  In the configuration example of the distance image sensor described above, four periods corresponding to the received light amounts A0, A1, A2, and A3 are set so that the phase interval is 90 degrees in one cycle of the modulation signal. If the phase with respect to the modulation signal is known, the four periods can be set at appropriate intervals other than 90 degrees. However, the formula for obtaining the phase difference ψ differs if the interval is different. In addition, the period of taking out the charge corresponding to the amount of received light in the four periods is within one period of the modulation signal as long as the reflectance of the object and the external light component do not change and the phase difference ψ does not change. It is not essential to take out four signal charges. Furthermore, when there is an influence of disturbance light such as sunlight or illumination light, it is desirable to arrange an optical filter that transmits only the wavelength of light emitted from the light source 5 in front of the photosensitive portion 3a. In the configuration example described with reference to FIGS. 13 and 14, three control electrodes 23 are associated with each photosensitive portion 3a. However, four or more control electrodes 23 may be provided. In the above example, the same configuration as the FT type CCD image sensor is adopted, but the same configuration as the interline transfer (IT) method and the frame interline transfer (FIT) method is adopted. Is also possible.

It is a block diagram of the abnormality detection apparatus of this embodiment. It is a block diagram of the distance image sensor part used for the same as the above. It is explanatory drawing explaining the setting method of a warning area same as the above. It is explanatory drawing explaining the other setting method of a warning area | region same as the above. It is explanatory drawing explaining operation | movement same as the above. (A)-(c) is explanatory drawing explaining other operation | movement same as the above. It is explanatory drawing explaining another operation | movement same as the above. It is explanatory drawing explaining another operation | movement same as the above. It is explanatory drawing explaining another operation | movement same as the above. (A)-(c) is operation | movement explanatory drawing of the photon detection element used for the same as the above. (A) (b) is operation | movement explanatory drawing of the principal part of the photon detection element used for the same as the above. It is a top view of the photon detection element used for the same as the above. (A) (b) is operation | movement explanatory drawing of the principal part of the photon detection element used for the same as the above. (A) (b) is operation | movement explanatory drawing of the principal part of the photon detection element used for the same as the above.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Abnormality detection apparatus 2 Distance image sensor part 10 Signal processing part 11 Image input means 12 Warning area setting means 13 Abnormality determination means 14 Alarm means

Claims (8)

A distance image sensor unit that images a target space and generates a distance image having a distance value as a pixel value;
A warning area setting unit that sets, as a warning area, an area that includes at least a warning object in the distance image input from the distance image sensor unit;
If there is a change in the image in the alert area in the distance image input from the distance image sensor unit, an abnormality determination unit that outputs a warning signal;
An abnormality detection device comprising: an alarm output unit that outputs an alarm when the alarm signal is input.   The warning area setting unit has means for detecting a plane that is a placement surface in the distance image, and determines an object placed on the detected plane as a warning object, and corresponds to the warning object The abnormality detection device according to claim 1, wherein an area to be operated is set as the warning area.   The alert area setting unit has means for designating a pixel corresponding to the alert object in the distance image, and sets an area in which pixel values are substantially continuous around the designated pixel as the alert area. The abnormality detection device according to claim 1, wherein:   The abnormality determination unit does not immediately output a warning signal when there is a change in the image of the alert area, and the image of the alert area changes before a predetermined time elapses after the change occurs. The abnormality detection device according to any one of claims 1 to 3, wherein the output of the alarm signal is stopped when the previous image is restored, and the alarm signal is output when the image does not return to the previous image.   The abnormality determination unit does not immediately output a warning signal when there is a change in the image of the alert area, and alerts if the ratio of the part where the shape of the alert object has not changed is equal to or greater than a certain ratio. The abnormality detection device according to any one of claims 1 to 4, wherein the output of a signal is stopped and an alarm signal is output if the ratio is less than a certain ratio.   The abnormality determination unit includes means for storing a distance image of a warning object imaged at normal time as a template image, and does not immediately output a notification signal when there is a change in the image of the warning area, Pattern matching is performed by relatively shifting the position of the template image with respect to the image of the alert area, and the determination condition is that the positional deviation when the correlation value is equal to or greater than a predetermined reference value is smaller than a predetermined threshold value. 6. The abnormality detection device according to claim 1, wherein if the condition is satisfied, the output of the alarm signal is stopped, and if the determination condition is not satisfied, the alarm signal is output.   The abnormality determining unit includes: a unit that stores a distance image of a warning object captured in a normal state as a template image; a unit that extracts edge pixels from the template image; and one or more blocks including at least a certain number of edge pixels. Means for extracting as a feature block, and when a change occurs in the image of the alert area, a warning signal is not immediately output, and the position of the feature block is shifted relative to the image of the alert area. Block matching is performed, and if the judgment condition that the positional deviation when the correlation value exceeds the predetermined reference value for each feature block becomes smaller than the predetermined threshold is satisfied, the output of the alarm signal is stopped. The abnormality detection device according to claim 1, wherein an alarm signal is output if the determination condition is not satisfied.   The abnormality detection device according to claim 1, wherein the warning area setting unit sets a range of a certain distance from the warning object as the warning area.

Images